MusicGen Music Generator

Generate original music compositions using Meta's MusicGen model with advanced features including audio continuation, multiple model versions, and precise sampling control for professional music production.

API Endpoint

POST /api/features

Parameters

Parameter	Type	Required	Description
type	string	Yes	Must be "MUSIC_GENERATOR"
model	string	Yes	Must be "meta/musicgen:671ac645ce5e552cc63a54a2bbff63fcf798043055d2dac5fc9e36a837eedcfb"
conversationId	string	No	Conversation ID for context
promptObject	object	Yes	MusicGen-specific parameters

Prompt Object Parameters

Parameter	Type	Required	Description
model_version	string	Yes	Model size: "stereo-melody-large", "stereo-large", "melody-large", "large"
prompt	string	Conditional	Text description (required if no audioUrl)
audioUrl	string	Conditional	Audio file URL for continuation (required if no prompt)
duration	number	Yes	Duration in seconds (1-30)
output_format	string	Yes	Output format: "mp3" or "wav"
normalization_strategy	string	Yes	Audio normalization: "loudness", "clip", "peak", "rms"
temperature	number	No	Sampling temperature (default: 1.0)
top_k	number	No	Top-k sampling (default: 250)
top_p	number	No	Top-p sampling (default: 0, disabled)
classifier_free_guidance	number	No	CFG scale (default: 3.0)
seed	number	No	Random seed for reproducibility
multi_band_diffusion	boolean	No	Enhanced audio quality (disabled for stereo models)
continuation	boolean	No	Enable audio continuation mode
continuation_start	number	Conditional	Start time for continuation (required if continuation is true)
continuation_end	number	Conditional	End time for continuation (-1 for full audio)

Model Versions

Version	Channels	Features	Best For
stereo-melody-large	Stereo	Melody conditioning, largest model	High-quality stereo with melody input
stereo-large	Stereo	Large model, stereo output	High-quality stereo music
melody-large	Mono	Melody conditioning, large model	Melody-guided generation
large	Mono	Standard large model	General music generation

Key Features

Audio Continuation

• Continue Existing Audio: Extend audio files seamlessly
• Melody Conditioning: Use audio as melodic reference without continuation
• Precise Control: Specify exact start/end points for continuation

Advanced Sampling

• Temperature: Control randomness and creativity
• Top-k/Top-p: Fine-tune token selection for quality vs diversity
• Classifier-Free Guidance: Balance prompt adherence vs creative freedom
• Seed Control: Reproducible results for consistent output

Interactive API Testing

API Playground

GETPOSTPUTDELETEhttps://api.1min.ai/api/features

Model Version *Stereo Melody Large - Best quality with melody conditioningStereo Large - High-quality stereo outputMelody Large - Mono with melody conditioningLarge - Standard high-quality mono

Music Description upbeat electronic music with synthesizer leads and driving bass

Duration (seconds) *

Output Format *MP3 - Compressed audio formatWAV - Uncompressed audio format

Normalization Strategy *Loudness - Perceptual loudness normalizationClip - Simple clipping normalizationPeak - Peak level normalizationRMS - Root mean square normalization

Temperature

Top K

Top P

Classifier Free Guidance

Seed

Audio Continuation No - Generate from text promptYes - Continue from audio file

Generated cURL Command:

curl -X POST "https://api.1min.ai/api/features" \
  -H "API-KEY: <your-api-key>" \
  -H "Content-Type: application/json" \
  -d '{
  "type": "MUSIC_GENERATOR",
  "model": "meta/musicgen:671ac645ce5e552cc63a54a2bbff63fcf798043055d2dac5fc9e36a837eedcfb",
  "conversationId": "MUSIC_GENERATOR",
  "promptObject": {
    "model_version": "stereo-melody-large",
    "prompt": "upbeat electronic music with synthesizer leads and driving bass",
    "duration": 10,
    "output_format": "mp3",
    "normalization_strategy": "loudness",
    "temperature": 1,
    "top_k": 250,
    "top_p": 0,
    "classifier_free_guidance": 3,
    "multi_band_diffusion": false,
    "continuation": false
  }
}'

Response Format

Successful requests return:

{}

Advanced Features

Multi Band Diffusion

• Enhanced Quality: Improves audio fidelity and reduces artifacts
• Compatibility: Only works with mono models (not stereo models)
• Trade-off: Longer generation time for better quality

Audio Continuation

• Seamless Extension: Continue existing audio files naturally
• Melody Conditioning: Use audio as melodic reference without continuation
• Precise Control: Specify exact segments for continuation

Sampling Parameters

Temperature (0.1-2.0)

• Low (0.1-0.7): More conservative, predictable results
• Medium (0.8-1.2): Balanced creativity and coherence
• High (1.3-2.0): More creative, less predictable results

Top-K Sampling (1-1000)

• Low (1-50): Very focused, limited vocabulary
• Medium (100-500): Balanced selection
• High (500+): More diverse, potentially chaotic

Classifier-Free Guidance (1-10)

• Low (1-3): More creative freedom, less prompt adherence
• Medium (3-5): Balanced prompt following
• High (6-10): Strict prompt adherence, less creativity

Best Practices

1. Model Selection: Use stereo models for final production, mono for experimentation
2. Duration Limits: Keep within 1-30 seconds for best quality
3. Prompt Clarity: Be specific about instruments, style, and mood
4. Parameter Tuning: Start with defaults, adjust based on results
5. Continuation Setup: Ensure audio files are clean and properly formatted

Example Requests

Electronic Music Generation

{
  "type": "MUSIC_GENERATOR",
  "model": "meta/musicgen:671ac645ce5e552cc63a54a2bbff63fcf798043055d2dac5fc9e36a837eedcfb",
  "promptObject": {
    "model_version": "stereo-melody-large",
    "prompt": "energetic electronic dance music with pulsing synthesizers, heavy bass drops, and crisp hi-hats",
    "duration": 15,
    "output_format": "wav",
    "normalization_strategy": "loudness",
    "temperature": 1.2,
    "top_k": 200,
    "classifier_free_guidance": 4
  }
}

Classical Music with High Quality

{
  "type": "MUSIC_GENERATOR",
  "model": "meta/musicgen:671ac645ce5e552cc63a54a2bbff63fcf798043055d2dac5fc9e36a837eedcfb",
  "promptObject": {
    "model_version": "large",
    "prompt": "elegant classical piano piece in minor key with emotional crescendos and delicate arpeggios",
    "duration": 20,
    "output_format": "wav",
    "normalization_strategy": "peak",
    "temperature": 0.8,
    "top_k": 150,
    "classifier_free_guidance": 3.5,
    "multi_band_diffusion": true
  }
}

Audio Continuation

{
  "type": "MUSIC_GENERATOR",
  "model": "meta/musicgen:671ac645ce5e552cc63a54a2bbff63fcf798043055d2dac5fc9e36a837eedcfb",
  "promptObject": {
    "model_version": "stereo-large",
    "audioUrl": "https://example.com/audio.wav",
    "continuation": true,
    "continuation_start": 10,
    "continuation_end": 20,
    "duration": 10,
    "output_format": "mp3",
    "normalization_strategy": "loudness",
    "temperature": 1.0,
    "classifier_free_guidance": 3
  }
}

Technical Specifications

Input Requirements

• Audio Files: WAV, MP3, M4A formats supported
• Sample Rate: Automatically resampled to model requirements
• Duration: Input audio can be any length, continuation segment specified

Output Specifications

• Sample Rate: 32kHz for all models
• Bit Depth: 16-bit for MP3, 32-bit float for WAV
• Channels: Mono or stereo depending on model version

Performance Considerations

• Generation Time: Varies by duration and model version
• Memory Usage: Larger models require more computational resources
• Multi Band Diffusion: Increases generation time by ~50%